There have been known vehicle periphery monitoring devices that detect a monitored subject that is present around an own vehicle from a captured image acquired using an on-board camera and notifies a driver that the monitored subject is present around the own vehicle.
In a vehicle periphery monitoring device of this type, the shape of an image portion of a subject (e.g., pedestrian) or the like that is detected from a captured image may change in accordance with a distance between a camera and the subject. For example, when a pedestrian is present at a position of a predetermined distance or more from the camera, a whole-body image of the pedestrian may be captured, while when a pedestrian is present within the predetermined distance from the camera, only the head of the pedestrian may be captured. In such a case, when a recognition process is performed on the pedestrian using feature quantities of the whole-body image of the pedestrian, there is a risk that an image portion of only the head is not recognized to be of the pedestrian. As a result, even though a monitored subject such as a pedestrian is present around an own vehicle, there is a risk that the driver is not notified of the fact.
In view of such a problem, there is proposed an image recognition device that performs, on an image portion of an object being a distinction target in a captured image, comparison of an image portion with a pattern of a specified part of a predetermined object (e.g., the head of a pedestrian), with respect to the image portion not having been recognized to be of the pedestrian in comparison with a pattern of the whole of the predetermined object (e.g., the whole body of a pedestrian) (e.g., see Patent Literature 1).